Selçuk Canbek

Fractal Dimension Based Shot Transition Detection in Sport Videos

Increase in application fields of video has boosted the demand to analyze and organize video libraries for efficient scene analysis and information retrieval. This paper addresses the detection of shot transitions, which plays a crucial role in scene analysis, using a novel method based on fractal dimension (FD) that carries information on roughness of image intensity surface and textural structure. The proposed method is tested on sport videos including soccer and tennis matches that contain considerable amount of abrupt and gradual shot transitions. Experimental results indicate that the FD based shot transition detection method offers promising performance with respect to pixel and histogram based methods available in the literature.

Gradual shot change detection in soccer videos via fractals

Accurate shot change detection plays vital role to organize video contents into meaningful parts for video scene analysis. Shot changes may occur with either abrupt or gradual shot transition. In abrupt shot transition, the change of video content occurs over a single frame. In gradual shot transition, however, the content change takes place gradually through a short sequence of frames. The gradual transition is also divided into several subgroups such as dissolve, pan, zoom and wipe. In the literature, there exist numerous methods to detect shot changes. While these methods are able to detect abrupt changes properly, they usually fail to detect gradual changes. This paper proposes a new method for detecting gradual shot changes successfully. The proposed method employs fractal dimension information of gray scale video frames. In the experimental work, efficacy of fractal based gradual shot change detection method is tested on soccer videos including different types of gradual shot changes and compared with well known pixel and histogram based shot change detection methods. Results of the experimental study indicate that the proposed method yields higher accuracy with respect to other methods in most cases.

Real-time Motion Region Detection and Contouring System

Many real world applications require real-time motion detection. Performance of a motion detection system should be fast enough so that moving objects in video can be detected and processed in real time. Once motion region in a video is detected, object tracking, image data mining, semantic meaning extraction, and other video/image processing techniques can be performed. There is a lot of research in realtime motion detection however most of these studies are either computationally expensive or not robust enough against real world image overheads like background clutter and pixel noise. In this paper, we propose a robust real-time motion detection system with low computational complexity. The proposed system finds the difference images from successive frames, applies filtering and the pixel clustering operation on them. Then, sub-contours of the observed motion regions are found using either minimum bounding coarse rectangles, minimum bounding rectangles using moments or minimum bounding convex hull methods. Finally, our contour merging algorithm merges sub-contours to identify moving object regions. This novel approach, detects moving objects in a video frame with tight enclosing boundary contours without sacrificing robustness. Since no constraint is attached to the technique, it is a good candidate to be used as motion detection of choice in wide range of applications. We obtained promising results in the experiments with real time data.

TASK ALLOCATION IN MARKET-BASED APPLICATIONS BY THE NEAREST NEIGHBOR HEURISTIC

One of the main subjects that is studied in multi-robot applications is allocating tasks to robots and constructing objective-efficient tours for the robots using these allocated tasks. The main purpose of this paper is to allocate tasks via market-based task allocation architecture and to construct collision-free routes for multi-robot systems in a known indoor environment. For task allocation, a market-based architecture is constructed and applied to an example of a team of four heterogeneous mobile robots. In the study, tasks are allocated according to one of two constraints: minimizing the makespan of the robots and maximizing the robot task matching value of each robot in the group. These constraints are achieved by changing the value of a decision parameter. To show the effectiveness of the proposed method, comparisons are made within these constraints. To construct path, a combination of the Nearest Neighbor heuristic and Dijkstras shortest path algorithms is used. To construct non-conflicting paths, a method that detects and solves collisions is developed. In the study, for agent communication Open Agent Architecture is used. Additionally, simulations on the MobileSim platform are conducted to verify the feasibility of the proposed method.

Modeling a Web Service-Based Decentralized Parallel Programming Environment

Abstract: Heterogeneous computer clusters which deploy COTS PCs and/or high performance server computers are economic parallel systems. However, development and execution software should be installed on each cluster host along with the operating system. Cluster members can no longer be used as standalone PCs without interrupting cluster work. On the other hand, usability of hosts in both individual and cluster-related tasks at the same time is strongly desired. In this paper, we present a web service-based parallel heterogeneous model in which participation is conditional and dynamic. Grid functionality is provided by using service oriented software architecture. Since service architecture is widely supported by operating systems, the hosts can easily be adapted to the parallel environment. Completion of all jobs is guaranteed theoretically and shown practically on the laboratory implementation as long as at least one host is operational in the system.

Plant classification using fractal dimension in HSV color space

In this paper, the use of fractal dimension measurements in HSV color space for image segmentation is investigated. The regions including different plant species, especially weed and non-weed, are analyzed as application. With this purpose, several features are extracted using HSV color components belonging to images, which contain different plant types, and corresponding partial fractal dimensions. Then, the extracted features are fed into Bayes classifier so that the regarding plant types are classified. Experimental results indicate that the features obtained with HSV components provide results as good as the ones that are extracted using gray scale values only. When all the features are used together, the best performance is obtained.

Detection of Partially Occluded Upper Body Pose Using Hidden Markov Model

This paper presents a method that assembles detected human body parts into a upper body human configuration. In the proposed method, six body parts (face, torso, legs and hands) are found by dedicated detectors using support vector machines (SVMs). Next, body configurations are assembled from the detected parts using hidden Markov model (HMM). Utilizing three different HMM with a decision mechanism, partially occluded human upper body poses are successfully assembled. The detection method shows promising results when tested on images from MIT pedestrian database and additional human body pictures.

Iterative super-resolution reconstruction using modified subgradient method

Modified subgradient method has been employed to solve super-resolution restoration problem. The technique uses augmented Lagrangians for nonconvex minimization problems with equality constraints. The subgradient of the constructed dual function is used for a measure. Initial results on comparative studies have shown that the technique is very promising.